xAI releases Grok 4.20
xAI released Grok 4.20 with a published system card; an earlier draft entry for "Grok 4.2 (Feb 17, 2026)" had no verifiable official release record and has been corrected to the official one.
SEARCH
Search English reports
English reports are listed here. AI-generated drafts are clearly labeled until an editor reviews them.
AI-generated · reviewedImpact 45/100
Google releases the Gemma 4 open model family
Google releases Gemma 4 for local inference, complex reasoning and agent workflows under the Apache 2.0 license.
OpenAI resets all Codex plans during unexplained usage spike
After Codex dashboards show an unexplained surge to usage caps, OpenAI removes abusive accounts and resets limits across every plan.
OpenAI resets Codex limits to promote new plugins
During a Codex plugin launch, OpenAI resets usage across every plan to encourage more users to try the new plugins.
Court pauses federal restrictions on Anthropic
The U.S. District Court for the Northern District of California issued a preliminary injunction pausing the supply-chain-risk finding against Anthropic and the related directives.
Report finds China's AI comic-drama boom becoming a production race
An industry report says AI comic-drama costs have fallen to hundreds of yuan per minute as competition shifts toward output, cost and hit rate.
Xiaomi launches MiMo V2 Pro for complex agent workflows
Xiaomi launches its flagship MiMo V2 Pro through the API, with a 1M-token context window and hybrid attention for complex agent workflows.
Xiaomi launches MiMo V2 Omni multimodal agent model
Xiaomi launches MiMo V2 Omni through the API, unifying text, vision and speech perception with tool execution and a 256K context window.
China's CNCERT warns of security risks in OpenClaw
China's national cyber emergency center flags prompt injection, unintended actions, malicious plugins and public exposure as OpenClaw risks.
Anthropic sues the Trump administration
Anthropic filed suit in the U.S. District Court for the Northern District of California, challenging the supply-chain-risk finding and the related federal restrictions.
OpenAI resets Codex limits after double-usage promotion failure
After about 9% of Plus and Pro users miss a double-usage promotion, OpenAI fixes the issue and resets Codex limits for both plans.
Trump orders federal agencies to stop using Anthropic technology
President Trump directed federal agencies to stop using Anthropic technology; the GSA then removed the products from USAi.gov and federal procurement catalogs.
OpenAI open-sources Symphony orchestration specification
Symphony turns project boards such as Linear into a control plane where coding agents claim tasks and deliver work from isolated environments.
Google introduces Nano Banana 2
Google positioned Nano Banana 2 as Gemini 3.1 Flash Image, for higher-quality, faster image generation and editing.
Block cuts more than 40% of staff and emphasizes AI workflows
Block announces workforce cuts exceeding 40%, linking the restructuring to organizational simplification and AI- and agent-based workflows.
Anthropic rejects the military's "any lawful use" clause
Anthropic refused to drop two restrictions - on large-scale domestic surveillance and fully autonomous weapons - causing its contract talks with the U.S. Department of War to break down publicly.
Nous Research releases Hermes Agent
Nous Research listed Hermes Agent as an agent project released on February 25, 2026, positioned as a long-running autonomous agent that runs on a server.
AgentScope team open-sources personal agent CoPaw
CoPaw joins the personal-agent ecosystem with local or cloud deployment, multiple chat integrations and an Apache 2.0 license.
Google DeepMind releases Gemini 3.1 Pro
A Pro-tier update in Google's 3.1 generation, further tuned for long-chain reasoning across modalities.
Google brings Lyria 3 music generation to Gemini
Google rolls Lyria 3 into Gemini, letting users create short songs with vocals from text prompts or uploaded images.